Genetic algorithms with self-organizing behaviour in dynamic environments

Abstract

In recent years, researchers from the genetic algorithm (GA) community have developed several approaches to enhance the performance of traditional GAs for dynamic optimization problems (DOPs). Among these approaches, one technique is to maintain the diversity of the population by inserting random immigrants into the population. This chapter investigates a self-organizing random immigrants scheme for GAs to address DOPs, where the worst individual and its next neighbours are replaced by random immigrants. In order to protect the newly introduced immigrants from being replaced by fitter individuals, they are placed in a subpopulation. In this way, individuals start to interact between themselves and, when the fitness of the individuals are close, one single replacement of an individual can affect a large number of individuals of the population in a chain reaction. The individuals in a subpopulation are not allowed to be replaced by individuals of the main population during the current chain reaction. The number of individuals in the subpopulation is given by the number of individuals created in the current chain reaction. It is important to observe that this simple approach can take the system to a self-organization behaviour, which can be useful for GAs in dynamic environments.